Improving LMP based Day Ahead Forecasts Using Auto Regressive Integrated Moving Average (ARIMA) with Shadow Pricing, EFORd Rates, and Transmission Loss Ratios

摘要

In this paper, forecasting methods for Day Ahead Locational Marginal Pricing (DA LMP) are investigated for PJM markets. Specifically, we explored the influence of parameters such as EFORd Rates, Shadow Pricing (SP) and Transmission Loss Ratios (TLR) in the data sets with forecasting models such as ARIMA, exponential smoothing, and Monte-Carlo methods (e.g., embedded in RT-Sim commercial software). Specifically, the approach used in this study is different from that of traditional methods as forecasting models are trained with EFORd, SP and TLR to investigate the effect of three LMP variables: Energy cost, Loss and Congestion. The methods relied on last one week of historical data to predict the day-ahead LMP, and its results were compared against PJM's forecasting methods for a known historic date. Our goal was to study the effect of three different parts that makeup LMP and use them rather than traditional trend pricing and weather variables to predict DA LMP pricing. Our preliminary findings indicate that using EFORd, SP and TLR for PJM's peak outage season coupled with conventional data appears to improve forecast accuracy nearly 9% (MAPE value) as compared to PJM's forecasting approach.